xyz1453


Total Posts: 4 
Joined: Aug 2014 


Maybe mutual Information and lempelziv complexity estimator can be helpful for when a signal is different than the expectations:https://arxiv.org/abs/1401.2548
Another suggestion might be finding the trading dependence with zscore https://books.google.com.tr/books?id=iUSzmtLJ3AgC&pg=PA8&lpg=PA8&dq=trading+Dependence+zscore&source=bl&ots=Ztgo5Eap9y&sig=YFBSCNgkBwrDBj6NN7NT70Kzvik&hl=en&ei=ZjKqTcibBcLpgAey6IT0BQ&sa=X&oi=book_result&ct=result&redir_esc=y#v=onepage&q=trading%20Dependence%20zscore&f=false 



rickyvic


Total Posts: 187 
Joined: Jul 2013 


Approaches are two: 1) use backtested returns and build a portfolio according to risk target, you can use anything including mean variance 2) use expected returns conditional mean and variance, possibly higher moments to build a portfolio.
It is a portfolio optimisation problem where you are looking to maximise returns or min risk (whatever estimate you want) with a number of constraints. If you use backtested returns from each strategy/instrument/prediction you are not surely using all the risk budget, if you are using the covariance matrix information you might be in the position of being wrong about correlations and over size risk.
Choice depends on uncertainty about predictions and your take on them. 
"amicus Plato sed magis amica Veritas" 
